Influence of Climate Regime Shift on the Interdecadal Change in Tropical. Cyclone Activity over the Pacific Basin during the Middle to Late 1990s

Manuscript Click here to download Manuscript clidy-d-15-00233-r3.docx Click here to view linked References 1 2 Influence of Climate Regime Shift on the Interdecadal Change in Tropical Cyclone Activity over the Pacific Basin during the Middle to Late 1990s 3 4 5 Chi-Cherng Hong 1, Yi-Kai Wu 12, and Tim Li 3 6 7 8 1 Department of Earth and Life, University of Taipei, Taipei, Taiwan 9 10 2 Department of Earth Sciences, National Taiwan Normal University, Taipei, Taiwan 3 IPRC/SOEST, University of Hawaii at Manoa 11 12 13 14 15 Submitted to Clim. Dyn. 16 17 18 19 20 21 22 23 Revised in January 2016 ------------------------------------------------------------------------------------------------------- Corresponding author: Dr. Chi-Cherng Hong, Department of Earth and Life, University of Taipei 1 Ai Kuo West Road, Taipei, Taiwan 10048. E-Mail: cchong@utaipei.edu.tw;hong0202@gmail.com FAX :886-2-23897641 1

24 Abstract 25 In this study, a new interpretation is proposed for the abrupt decrease in tropical 26 cyclone (TC) activity in the western North Pacific (WNP) after the late 1990s. We 27 hypothesize that this abrupt change constitutes a part of the phenomenon of 28 interdecadal change in TC activity in the Pacific Basin, including the WNP, western 29 South Pacific (WSP), and eastern North Pacific. Our analysis revealed that the 30 climate- regime shift (CRS) in the Pacific during the middle to late 1990s resulted in a 31 La Niña-like mean state, which was responsible for the interdecadal change in TC 32 activity in the late 1990s. Analyses of the TC genesis potential index and numerical 33 experiments revealed that the decline in TC activity in both the WNP and WSP was 34 primarily attributable to the increase of vertical wind shear in the central Pacific due 35 to the La Niña-like associated cold sea surface temperature (SST). Conversely, the La 36 Niña-like associated warm SST in the western Pacific produced anomalous vertical 37 transport of water vapor, increasing moisture levels in the mid-troposphere and TC 38 activity in the western WNP. Furthermore, the CRS modified the mean TC genesis 39 position and shifted the steering flow to the west, resulting in the increased frequency 40 of TC landfalls in Taiwan, southeastern China, and northern Australia after the late 41 1990s. 2

42 1. Introduction 43 The western North Pacific (WNP) constitutes part of the warm pool, where tropical 44 cyclone (TC) genesis occurs the most frequently. TC activity in the WNP exhibits 45 marked interannual and interdecadal fluctuation (Chan 2000; Chia and Ropelewski 46 2002; Wang and Chan 2002; Camargo and Sobel 2005; Chan 2005). For example, the 47 frequency of TC genesis was unusually high in 1994, when 34 TCs formed; however, 48 TCs numbered only 26 in 1995. Broadly, TC genesis frequency was high in the 49 mid-1960s and mid-1990s, but relatively low in the 1980s and late 1990s (Liu and 50 Chan 2013; Choi et al. 2015; He et al. 2015). Interannual variation in TC activity in 51 the WNP is associated with large-scale circulation anomalies caused by the El 52 Niño Southern Oscillation (Li 2012) and the stratospheric quasi-biennial oscillation 53 (Chan 2005; Liu and Chan 2013). Conversely, the strength and extent of the North 54 Pacific subtropical high (NPSH) are the principal factors modifying the interdecadal 55 variation in TC activity in the WNP (Chan 2005; Hsu et al. 2014). 56 The tropical Pacific underwent a basin-scale climate regime shift (CRS) during the 57 middle to late 1990s (Hong et al. 2013). This CRS, which was characterized by a 58 negative sea surface temperature anomaly (SSTA) in the equatorial eastern Pacific and 59 a positive SSTA in both the tropical western Pacific and the entire subtropical Pacific, 60 was accompanied by a pair of low-level atmospheric anticyclone and cyclone 3

61 anomalies in the tropical western and eastern Pacific. The low-level anticyclone 62 anomaly in the WNP facilitated the NPSH to extend westward, potentially 63 suppressing TC activity in the WNP (Chia and Ropelewski 2002; Liu and Chan 2013; 64 Chan 2005). 65 Recent studies have identified a significant interdecadal decrease in TC activity in 66 the WNP after the late 1990s (Liu and Chan 2013; Hsu et al. 2014; and He et al. 2015). 67 These studies attributed this reduction to increases in both the large-scale vertical 68 shear and the 500-hPa geopotential height. Although the decline of TC activity in the 69 WNP after the late 1990s has been investigated extensively, most previous studies 70 have focused on the specific region of TC genesis in the WNP. Thus, it remains 71 unclear whether this interdecadal change can be detected in adjacent genesis regions, 72 such as the western South Pacific (WSP) and eastern North Pacific (ENP). 73 Additionally, the physical processes underlying the interdecadal increase in TC 74 genesis frequency in the northwestern part of the WNP remain obscure (He et al. 75 2015). The simultaneous occurrence of both the abrupt change in TC activity and the 76 CRS in the Pacific during the middle to late 1990s suggests that they were related. 77 Therefore, this study considered the role of the basin scale CRS in the abrupt change 78 in TC activity in the late 1990s. We hypothesized that the abrupt decrease in TC 79 activity in the WNP was associated with the greater phenomenon of interdecadal 4

80 change in TC activity in the Pacific Basin. This includes changes in TC activity in the 81 WNP and the WSP, and a westward shift of the mean TC genesis position in the WNP 82 and the ENP. In addition, the large-scale factors causing the interdecadal change in TC 83 activity in the Pacific were related to La Niña-like background (mean) changes, which 84 ultimately resulted from the Pacific CRS during the middle to late 1990s. 85 86 2. Data and Methodology 87 Data from the International Best Track Archive for Climate Stewardship (Knapp et 88 al. 2010) for the period 1960 2012 were used to analyze the change in TC frequency. 89 This data set was obtained from the Regional Specialized Meteorological Center of 90 the World Meteorological Organization and other agencies (e.g., the Joint Typhoon 91 Warning Center and the Japan Meteorology Agency). Because tropical depression 92 wind speeds in the Southern Hemisphere were not available until 1980, TC activity in 93 the WSP was analyzed using data starting from that year. In addition, the typhoon 94 season (October April) covers two calendar years in the Southern Hemisphere. 95 Therefore, the annual mean number of TCs within this region was based on the period 96 from July to the following June. Monthly atmospheric data sets based on the 97 NCEP/NCAR Reanalysis I (Kalnay et al. 1996) and monthly SST fields from the Met 98 Office Hadley Centre sea ice and sea surface temperature data sets (Rayner et al. 2003) 5

99 were employed for diagnosing the large-scale circulation anomalies. 100 A regime-shift index (RSI) (Rodionov 2004) was used to detect the specific time of 101 the CRS. Here, a cutoff length of 10 years was applied for CRS diagnosis, and only 102 values of the RSI that exceeded the 0.05 significance level were considered. A 103 sensitivity test of the CRS was performed by calculating the RSI with cutoff lengths 104 of 8 and 12 years (not shown). The test indicated that the abrupt change in TC activity 105 in the Pacific in the late 1990s was robust. 106 107 3. Abrupt Change in TC Activity in the Pacific Basin in the Late 1990s 108 3.1 TC genesis frequency 109 Figure 1 illustrates a time series of the annual TC count in various genesis regions 110 of the Pacific. The number of TCs in the WNP demonstrated clear interdecadal 111 variation, with active periods in 1965 1975 and 1985 1995, and inactive periods in 112 1976 1984 and 1998 2012. The RSI calculation shows that during the 53-year period 113 (1960 2012), three significant and abrupt changes occurred in the number of TCs in 114 1968, 1988, and 1998. Among these years, the abrupt decrease in TC activity in 1998 115 was particularly pronounced. The average annual number of TCs was 29 during the 116 active period of 1984 1997; however, it decreased to 24 during the inactive period of 117 1998 2011. 6

118 The time series of TC activity in the WSP illustrates that TC genesis frequency was 119 high in the 1990s and low in the 2000s. The RSI calculation indicates that the number 120 of TCs decreased abruptly and significantly in 1999. Because the typhoon season 121 (October April) covers two calendar years in the Southern Hemisphere, this change in 122 TC activity can be considered concurrent with the decrease in TC activity in the WNP 123 in 1998 on the interdecadal time scale. Further examination showed that the abrupt 124 changes in TC activity in the WNP and the WSP during the late 1990s were caused 125 primarily by a change in TC activity during the typhoon season (i.e., June to October 126 in the Northern Hemisphere, and November to the following April in the Southern 127 Hemisphere). TC activity during the nontyphoon season did not exhibit any 128 significant interdecadal change. The sensitivity of the RSI for 1998 1999 was further 129 tested using cutoff lengths of 8 and 12 years, and the results demonstrated that the 130 change point for the period of approximately 1998 1999 was robust. Considering this 131 change point, the active period of TC genesis, 1986 1997, was denoted as IP1, and 132 the inactive period, 1998 2009, as IP2. 133 The TC activity in the ENP reveals that TC genesis frequency was high during 134 1980 1994, however, it did not show significant change after 1995. Weak TCs (i.e., 135 tropical storms, having a maximum sustained wind speed of 34 63 knots) exhibited 136 an interdecadal increase (Fig. 1d), whereas strong TCs (having a maximum sustained 7

137 wind speed of > 63 knots) exhibited an interdecadal decrease after the late 1990s (i.e., 138 the result of Fig. 1c subtracted from that of Fig. 1d). This reversed interdecadal 139 change is similar to that described by Lupo et al. (2011). Because the interdecadal 140 increase of weak TCs was offset by the interdecadal decrease of strong TCs, the total 141 annual mean number of TCs in the ENP did not show a significant interdecadal 142 change in the late 1990s. 143 Figure 2 presents the horizontal distribution of the annual mean and variation of the 144 frequency of TC genesis in the Pacific during IP1 and IP2. The most remarkable 145 changes were the decrease in TC activity in the eastern WNP (east of 145 E) and WSP, 146 and the increase in TC activity in the western WNP (west of 145 E). Because the 147 decrease in the number of TCs in the eastern WNP was more pronounced than the 148 increase in the western WNP, the total number of TCs in the WNP exhibited a decline 149 after the late 1990s. The decrease and increase in TC activity in the eastern and 150 western WNP, respectively, caused the mean position of TC genesis in IP2 to shift 151 westward significantly from 149 E to 140 E (Table 1). Recent studies have shown 152 that the zonal asymmetry of local SST in the WNP (Choi et al. 2015) and the 153 tropical tropospheric warming (Wu et al. 2015) might lead to interdecadal change in 154 the westward shift of the mean TC genesis position in the WNP. Through 155 numerical simulations, we demonstrated that the La Niña-like associated negative 8

156 SSTA in the central Pacific might also facilitate a westward shift of the mean TC 157 genesis position. As in the WNP, a similar westward shift of the mean TC genesis 158 position from approximately 95 W to 102 W was detected in the ENP (Fig. 2c). This 159 westward shift led to a change in TC genesis frequency, suggesting a response to 160 global warming (Murakami et al. 2013). This indicates that the westward shift of the 161 mean TC genesis position in the ENP was not accounted for as a specific factor. In 162 contrast to the WNP, the interdecadal change of TC genesis frequency in the WSP 163 between IP2 and IP1 had an N S dipole-like structure. In other words, TC genesis 164 frequency decreased and increased in the northern and southern WSP, respectively 165 (Fig. 2c). 166 3.2 TC genesis track 167 Figure 3 depicts the spatial distribution of the mean TC track density in IP1 and IP2. 168 The TC track density exhibited a remarkable decrease in the southern WNP (south of 169 20 N), southeastern WSP (approximately 150 180 E), and southern ENP (south of 170 10 N) in IP2, among which the decrease in the southeastern WNP was the most 171 pronounced. Conversely, an increase in TC track density was observed for particular 172 regions, namely the region around Taiwan, the southeastern South China Sea, 173 northeastern Australia, and the Gulf of California. The increase of TC track density in 174 the WNP and WSP can be attributed to the substantial increase in TC genesis 9

175 frequency in the eastern Philippines (120 135 E) in IP2, owing to the westward shift 176 of the mean TC genesis position (Fig. 2c). In addition, the westward steering flow was 177 strengthened, experiencing a northward shift near the Philippine Sea in IP2 (Fig. 3c). 178 This interdecadal change in steering flow, combined with the westward shift of the 179 mean TC genesis position, led to a greater frequency of TC landfalls in Taiwan and 180 southeastern China after the late 1990s. 181 The increase of TC track density in northern Australia can also be explained by the 182 combined effects of the interdecadal change in steering flow and TC genesis 183 frequency. The strengthening and westward extension of the easterly steering flow in 184 IP2, together with the increase in TC genesis frequency off the coast of northeastern 185 Australia (Figs. 2c and 3c), led to an increase in TC track density in northern Australia, 186 which corresponds to one of the major TC tracks of Cluster 5 (Ramsay et al. 2012; Fig. 187 2), where TCs that formed in the South Solomon Sea moved westward toward the 188 coastline of northwestern Australia. This increase in TC track density was associated 189 with a remarkable increase in the number of severe TCs making landfall over eastern 190 Australia after the late 1990s (Callaghan and Power 2010) (figure not shown). 191 Another notable increase in TC track density occurred in northwestern Australia, 192 where the TC tracks corresponded to those of Cluster 4 (Ramsay et al. 2012). The 193 increase in TC track density was evidently caused primarily by the increase in TC 10

194 genesis frequency off the coast of northwestern Australia (Fig. 2c), where rapid 195 warming was identified after the 1980s (not shown). In the ENP, the decrease in TC 196 track density was primarily caused by the decrease in TC genesis frequency. Thus, 197 Figs. 2 and 3 clearly demonstrate that the interdecadal decrease in TC activity in the 198 WNP after the late 1990s (Liu and Chan 2013; Choi et al. 2015; He 2015) was part of 199 the phenomenon of basin-scale change in TC activity in the Pacific. 200 A comparison of annual mean numbers of TCs in various genesis regions of the 201 Pacific Basin between IP1 and IP2 is provided in Table 2. Decreases in TC activity in 202 both the WNP and WSP in IP2 are evident; the annual mean number of TCs in the 203 WNP and WSP decreased by approximately 17% and 20%, respectively. The decrease 204 in TC activity in the eastern WNP was partially offset by the increase in TC activity in 205 the western WNP. Regarding the specific genesis region of the eastern WNP, the 206 number of TCs decreased by approximately 60% in IP2. The TC activity in the ENP 207 also decreased (approximately 13%) in IP2, although the change was nonsignificant. 208 Irrespective of TC genesis frequency, the average lifetime of TCs also experienced a 209 remarkable decrease, particularly in the WNP, from IP1 to IP2 (Table 3). The most 210 pronounced change occurred in the WNP, where the average TC lifetime of 211 approximately 10 days in IP1 decreased to 8 days in IP2. A previous study suggested 212 that this shortening of TC lifetime could have led to the recent significant decrease in 11

213 the typhoon destructive potential index, despite the increase in TC intensity as a result 214 of global warming (Lin et al. 2015). 215 216 4. Causes of Changes in TC Activity 217 4.1 Mean-state change 218 Figure 4 shows the changes in the background condition between IP1 and IP2. 219 There is an evident La Niña-like or central Pacific (CP)-type La Niña (Ashok et al. 220 2007) SSTA pattern, accompanied by cooling and warming in the equatorial central 221 and western Pacific, respectively, and easterly wind anomalies (i.e., strengthened trade 222 winds in both hemispheres) in the tropical western Pacific (Fig. 4a). In association 223 with the K-shaped warm SST pattern in the western Pacific, an increase in relative 224 humidity in the mid-troposphere was identified (Fig. 4b), and the increases in relative 225 humidity along the South Pacific Convergence Zone (SPCZ) were particularly 226 pronounced (Fig. 4b). Relative humidity is defined as q/ q s, where q and q s 227 denote the specific humidity and statured specific humidity, respectively. Because 228 both q and q s increase simultaneously in response to warming, the increase of 229 relative humidity in the mid-troposphere is not necessarily attributable directly to SST 230 warming. A moisture budget revealed that the increase in relative humidity at 700 hpa 231 resulted primarily from the vertical climatological mean vapor transport caused by the 12

232 anomalous vertical velocity ( w' q). Specifically, the K-shaped warm SST pattern 233 led to an anomalous vertical velocity (Fig. 5), which transported low-level humid air 234 into the mid-troposphere, resulting in an increase in moisture at 700 hpa. Conversely, 235 the 850 hpa vorticity exhibited a remarkable decrease in the CP in IP2 (Fig. 4c). This 236 decrease was evidently caused by a pair of anomalous anticyclones in the CP (i.e., a 237 Gill-type response to the CP cold SST). These anomalous anticyclones produced 238 easterly anomalies in the equatorial CP, increasing the vertical wind shear (200 hpa 239 minus 850 hpa; Fig. 4d). Matsuura et al. (2003) described the influence of dynamic 240 conditions on interdecadal change in TC activity in the WNP, and other recent studies 241 (e.g., Liu and Chan 2013; Hsu et al. 2014; He et al. 2015) have emphasized further 242 that changes in dynamics were responsible for the interdecadal decrease in the WNP 243 after the late 1990s. The findings of this study are in accordance with these earlier 244 studies regarding the interdecadal decrease of TC activity in the eastern WNP after the 245 late 1990s. In addition, we demonstrated that thermodynamics, rather than dynamics, 246 have contributed to the interdecadal increase of TC genesis in the western WNP. The 247 role of thermodynamics in TC activity in the South China Sea was also mentioned by 248 Yang et al. (2012). The increase in TC genesis frequency in the northwestern ENP was 249 primarily caused by a low-level anomalous cyclone. Additionally, the cyclone 250 anomaly resulted in an anomalous upward motion. This transported a remarkable 13

251 quantity of moist air from the boundary layer to the mid-troposphere ( w' q 0), 252 providing a favorable environment for TC genesis. Conversely, the influence of 253 large-scale factors on the TC activity decrease in the southeastern ENP was 254 insignificant, a phenomenon reflected in the discrepancy between the changes in the 255 genesis potential index (GPI) and TC genesis frequency (Fig. 2c, Fig. 6). This 256 suggests that other physical processes (such as the intraseasonal oscillation), rather 257 than large-scale factors, contribute substantially to the interdecadal decrease in TC 258 activity in the southeastern ENP. 259 4.2 GPI analysis 260 Section 4.1 described the association between the change in background conditions 261 and the change in TC activity after the late 1990s. In this section, we explain 262 quantitatively how the mean-state change affected TC activity according to GPI 263 analysis. The GPI (Emanuel and Nolan 2004; Camargo et al. 2007) was applied to 264 examine the specific dynamic and thermodynamic factors that caused the abrupt 265 decline in TC activity in the late 1990s. The GPI is defined as follows: 266 267 (1) 268 (A) (B) (C) (D), 269 where denotes the absolute vorticity, H is the relative humidity at 700 hpa, v pot 14

270 is the potential intensity, and V shear is the vertical shear of the magnitude of the 271 horizontal wind between 200 and 850 hpa. Terms A and D are primarily related to the 272 background dynamic conditions, whereas Terms B and C are associated with the 273 thermodynamic conditions. 274 Figure 6 depicts the spatial distribution of the difference in the GPI between IP1 275 and IP2 (i.e., IP2 minus IP1). Compared Fig. 6 with Fig. 2, the decrease in TC activity 276 in the eastern WNP and WSP, as well as the increase in TC activity in the western 277 WNP and northwestern ENP, are well represented by the GPI. Considering this 278 outcome, we examined the contributions of the individual terms. Here, the 279 contribution of each term to the total difference in the GPI was calculated by 280 following the suggestions of Li et al. (2013). For example, the term δa B C D 281 denotes a change caused by Term A, whereas the other terms remain unchanged; δa 282 denotes the difference in Term A between IP1 and IP2. 283 Five regions (indicated by boxes in Fig. 6) where the TC genesis frequency 284 exhibited significant change were analyzed. The analysis revealed that the abrupt 285 decrease in the GPI in the eastern WNP during the late 1990s was caused primarily by 286 Terms A and D, which represent changes in the background dynamics (vorticity and 287 vertical wind shear). Conversely, the change in the thermodynamics (relative humidity 288 at 700 hpa) contributed to the increase in TC activity in the western WNP. In the ENP, 15

289 the increase in TC activity resulted primarily from changes in vorticity (Term A) and 290 vertical wind shear (Term D), with a partial contribution from relative humidity (Term 291 B). In the WSP, the decrease in the northern WSP was caused primarily by changes in 292 vorticity (Term A) and a partial contribution from vertical wind shear (Term D). 293 Conversely, the increase of GPI in the southern WSP resulted from changes in relative 294 humidity (Term C). Overall, the dynamic factors regulated the interdecadal decrease 295 of the GPI in the WNP and WSP, whereas the interdecadal increase in the GPI in the 296 western WNP (southeastern South China Sea and Philippine Sea) off the coast of 297 northeastern Australia and the SPCZ was caused primarily by the K-shaped warm 298 SST pattern associated with thermodynamic changes. 299 300 5. Influence of Basin-Scale CRS on the TC Activity 301 Figure 4 shows that the abrupt change in TC activity in the Pacific after the late 302 1990s was associated with a La Niña-like background change. Thus, it is reasonable 303 to consider what caused the change in background conditions. Figure 4 indicates that 304 both the decrease in the local relative vorticity and the increase in local westerly shear 305 were related to the anomalous low-level easterlies in the equatorial western Pacific. 306 The anomaly of the low-level easterlies or westward shift in the vertical wind shear 307 was caused by the change in the zonal SST gradient, which was ultimately associated 16

308 with the CRS in the Pacific Basin, as discussed by Hong et al. (2013). 309 Figure 7 depicts the first two leading singular value decomposition (SVD) modes of 310 the interdecadal SST and 850 hpa streamfunction fields derived from Hong et al. 311 (2013). Whereas the first SVD mode, which has a pattern resembling the Pacific 312 decadal oscillation (PDO) (pattern correlation coefficient: 0.95), exhibited a 313 significant abrupt change in 1976 1977, the second SVD mode showed a significant 314 abrupt change in 1996 1997. These changes were characterized by warming in the 315 tropical western Pacific and cooling in the equatorial CP. Accompanying the SST 316 pattern was a Matsuno Gill (Gill 1980) circulation pattern, with a pair of cyclonic and 317 anticyclonic circulation anomalies in the equatorial eastern and western Pacific, 318 respectively (Fig. 7). The mean-state change illustrated in Fig. 5 resembles that of the 319 SVD2 (pattern correlation coefficient: 0.91); the SSTA (with maximum observed 320 negative SSTA in the CP) is coherent with the quadratic circulation anomalies of an 321 anticyclone and cyclone in the western and eastern Pacific, respectively. The 322 similarity between the spatial pattern and temporal change point suggests that the 323 abrupt decrease in TC activity in the mid-1990s was caused by the basin-scale CRS in 324 the Pacific during 1996 1997. 325 Liu and Chan (2013) argued that the significant decrease in TC activity in the WNP 326 after the late 1990s was related to the PDO-like or interdecadal Pacific oscillation 17

327 (IPO) (Meheel et al. 2013) SST change. However, the RSI calculation demonstrated 328 that the PDO and IPO experienced a regime shift approximately in 2003 and 2005 329 respectively (figure not shown),lagging behind the CRS in 1996 1997 by 330 approximately 6 8 years. This indicates that the initiation of the change in 331 background conditions in the late 1990s resulted from SVD2, but not the PDO. 332 Further examination revealed that the abrupt change of the SVD2-associated second 333 principle component (PC2) in the middle to late 1990s was independent of the trend 334 of global warming, even though that trend might have enhanced the magnitude of PC2 335 (figure not shown). 336 The basin-wide SST underwent a significant regime shift in approximately 337 1996 1997. Before this regime shift, a warm SSTA appeared in the equatorial CP. In 338 response to this warm anomaly, a Matsuno Gill-type circulation anomaly appeared 339 with a low-level cyclonic anomaly in the WNP. This large-scale cyclonic circulation 340 anomaly, together with reduced vertical shear, favored TC genesis, accounting for 341 active TCs in the eastern WNP and the WSP in IP1. However, these anomalous 342 oceanic and atmospheric conditions reversed after 1996 1997. The SSTA in the 343 equatorial CP changed from positive to negative, whereas the SSTA in the equatorial 344 western Pacific changed from negative to positive. Because of the SSTA phase change 345 in the central equatorial Pacific, a low-level anticyclone anomaly occurred in the 18

346 WNP during IP2. Meanwhile, the vertical shear increased in response to the low-level 347 anticyclone anomaly. Both of these changes in the background state suppressed TC 348 activity in the eastern WNP during IP2. Conversely, in response to the K-shaped warm 349 SST pattern, an anomalous cyclone and enhanced vertical motion occurred in the 350 western WNP. As shown in Fig. 4, these thermodynamic changes resulted in the 351 interdecadal increase in TC genesis in the western WNP. 352 The influence of the basin-scale CRS on the changes in TC activity in the Pacific 353 was demonstrated further using regression analysis. Figure 8a shows the GPI 354 regressed on PC2. Compared with the observed GPI difference between IP1 and IP2 355 (Fig. 5), the negative GPI anomaly in the CP, as well as the positive GPI anomaly in 356 the western Pacific and ENP, were captured successfully by the regression map 357 (pattern correlation coefficient: 0.85). Further calculation showed that the regression 358 of the dynamics (vertical wind shear) and thermodynamics (700 hpa relative humidity) 359 for PC2 resembled the mean-state change between IP1 and IP2 (Fig. 8b and 8c). 360 These results reinforce the suggestion that the SVD2-induced La Niña-like 361 background change resulted in the interdecadal change of TC activity in the Pacific 362 after the late 1990s. 363 364 6. Numerical Experiments 19

365 The regression analysis revealed that the CRS influenced the change in TC activity 366 after the 1990s. The CRS-related SVD2 was characterized by a K-shaped positive 367 SSTA in the western Pacific and a negative SSTA in the central eastern Pacific. In 368 this section, we describe the influence of the SVD2-related SSTA on the interdecadal 369 change in TC activity in the Pacific according to numerical experiments. The effects 370 of the SVD2-related K-shaped warming and CP-type cooling on the GPI are also 371 discussed. 372 A general circulation atmospheric model called the simplified parameterizations 373 primitive-equation dynamics (SPEEDY) model (Molteni 2003) was employed in the 374 numerical experiments. The SPEEDY model is noted for its efficiency and has been 375 used successfully in several other studies related to monsoonal dynamics (Bracco et al. 376 2005) and the interdecadal variation of large-scale circulations (Kucharski et al. 2006). 377 Table 4 presents the details of the numerical experiments designed for this study. In 378 the control experiment (CTL), the model was forced by the observed climatological 379 monthly SST during the period 1960 2010. In the three major experiments, namely 380 PA, WPA, and EPA, the SVD2-related SSTA, positive SSTA of SVD2 (K-shaped 381 warming), and negative SSTA of SVD2 were respectively prescribed to the model. 382 Ten-member ensemble runs were performed for each experiment. Subsequently, each 383 experiment was integrated for 20 years, and the 20-year mean of the SST-forced 20

384 atmospheric circulations was assumed as the background mean state. 385 Figure 9b illustrates the anomalous 850 hpa streamfunction in response to the 386 SSTA between IP1 and IP2. Comparing observations (Fig. 9a) revealed that the pair of 387 anticyclonic and cyclonic circulation anomalies in the western and eastern Pacific, 388 respectively, was replicated successfully. The anomalous cyclonic circulation in the 389 South China Sea was also captured. The coefficient of the spatial pattern of the 850 390 hpa streamfunction between the observations and simulation reached 0.83. This 391 similarity indicates that the atmospheric mean-state change during the late 1990s in 392 the Pacific was forced by the SVD2-related SSTA. 393 Figure 9c and 9d shows the steady response of the 850 hpa streamfunction to the 394 CP-like cooling (experiment EPA) and K-shaped warming (experiment WPA), 395 respectively. The response revealed that the pair of low-level anticyclonic and 396 cyclonic anomalies in the western and eastern Pacific, respectively, was forced 397 primarily by the associated CP-like negative SSTA. Conversely, the cyclonic 398 circulation anomaly in the South China Sea was maintained by the positive SSTA in 399 the western Pacific. Experiment WPA (Fig. 9d) also indicated that the positive SSTA 400 in the western Pacific might have act to enhance the equatorial zonal SST gradient, 401 which would be strengthen the easterly anomaly (i.e., an enhancement of the 402 prevailing trade wind) in the CP (near 180 160 W); consequently, the vertical wind 21

403 shear in the CP was enhanced. Additionally, the enhancement of the prevailing trade 404 winds in the Pacific would produce positive feedback for the La Niña-like SSTA 405 (Matthew et al. 2014). 406 Finally, the change in the GPI in response to each SSTA (Table 3) is shown in Fig. 407 10b d. Although the negative GPI in the CP was overestimated and had an eastward 408 extension, the large-scale change in the GPI between IP1 and IP2 was accurately 409 simulated. The success of experiment PA in simulating the GPI change again indicated 410 that the decrease in TC activity in the Pacific basin was caused primarily by the 411 CRS-related SSTA. The contributions of experiments EPA and WPA to the total GPI 412 change are shown in Fig. 10b and 10c, respectively. The figure illustrates that the 413 negative changes in the GPI in the eastern WNP and the WSP were caused primarily 414 by CP-type cooling. Conversely, the positive change in the GPI in the Philippine Sea 415 was caused by the associated K-shaped warming in the western Pacific. Additionally, 416 the zonally distributed positive changes in the GPI in the subtropics of both 417 hemispheres were caused primarily by the positive SSTA in the western Pacific. 418 Figure 10c also suggests that the northward shift of the TC genesis mean position in 419 the WNP was a result of the K-shaped warming. Thus, the numerical experiments 420 confirmed the findings of the regression analysis and demonstrated that the large-scale 421 decline in TC activity in the eastern WNP and northern WSP after the late 1990s was 22

422 attributed to CP-type cooling. Nevertheless, the increase in TC activity in the western 423 WNP and southern WSP (distributed along the SPCZ) was caused primarily by 424 K-shaped warming of the SST in the western Pacific. 425 426 7. Conclusions and Discussion 427 This study examined whether the abrupt change in TC activity in the Pacific in the 428 middle to late 1990s was caused by changes in the background conditions induced by 429 the basin-scale CRS in the middle to late 1990s, and how the basin-scale CRS affected 430 the interdecadal change in TC activity in the Pacific through GPI analysis and 431 numerical experiments. It has been established that a basin-scale CRS in the Pacific 432 and an abrupt change in TC activity in various genesis regions, including the WNP, 433 WSP, and ENP, occurred simultaneously in the middle to late 1990s. The 434 simultaneous concurrence of these two abrupt changes suggests that they were 435 possibly related. We hypothesized that the abrupt change in TC activity in the Pacific 436 Basin during the late 1990s was primarily caused by the CRS-induced La Niña-like 437 SSTA pattern. The basin-scale CRS after the late 1990s induced large-scale changes in 438 both oceanic and atmospheric conditions, which led to the abrupt decline in TC 439 genesis in the WNP and a significant change in TC activity in the ENP and WSP. The 440 main results of this study are summarized as follows: 23

441 442 1. Significant changes in TC activity were identified during the late 1990s in various 443 genesis regions of the Pacific. These included a marked decrease in TC activity in the 444 eastern WNP and the WSP, and an increase in TC activity in the Philippine Sea and 445 southeastern South China Sea. Irrespective of the interdecadal change in the number 446 of TCs, the mean TC genesis position in the WNP and ENP also exhibited a 447 remarkable westward shift after the late 1990s. 448 2. A GPI diagnosis revealed that the major large-scale factors that caused the decrease 449 in TC activity in the eastern WNP and northern WSP were related to changes in the 450 dynamic conditions (i.e., a decrease in low-level vorticity and an increase in vertical 451 wind shear). Conversely, the change in thermodynamic conditions (i.e., increase in 452 700 hpa relative humidity) contributed to the abrupt increase in TC activity in the 453 western WNP and southern WSP. The increase in 700 hpa relative humidity was 454 caused by the enhancement of vertical climatological mean vapor transport by the 455 anomalous vertical velocity produced by the K-shaped warming. Both the 456 thermodynamic and the dynamic conditions contributed to the change of TC activity 457 in the ENP. 458 3. Observations revealed that the changes in the background conditions from IP1 to 459 IP2 were caused primarily by the basin-scale CRS during the middle to late 1990s. A 24

460 GPI regression in the CRS-related PC2 confirmed that the CRS could have led to the 461 interdecadal change in TC activity in the Pacific in the late 1990s. 462 4. Numerical experiments showed that the interdecadal decline in TC genesis in the 463 eastern WNP and northern WSP was primarily attributable to the CRS-related CP-type 464 cooling. However, the CRS-related K-shaped warming in the western Pacific 465 contributed to the interdecadal increase in TC genesis in the western WNP and 466 southern WSP. 467 468 It has been identified that an abrupt change occurred in TC activity in the various 469 genesis regions of the Pacific during the late 1990s. The interdecadal decrease in TC 470 activity in the WNP during the late 1990s has been widely reported in recent studies 471 (Liu and Chen 2013; Hsu et al. 2014; Choi et al. 2015; He et al. 2015). However, in 472 this paper, a new interpretation of this interdecadal change is proposed. Furthermore, 473 several new scientific findings are illustrated, such as the decrease in TC genesis 474 frequency in the northern WSP, the physical processes leading to the increase in TC 475 genesis frequency in the western WNP, and the increase of TC landfalls over Taiwan 476 and northeastern Australia. Here, we assumed that the abrupt change in TC activity in 477 the WNP during the late 1990s was part of the phenomenon of basin-scale change in 478 TC activity in the Pacific (i.e., a decrease in TC activity in the eastern WNP and 479 northern WSP, an increase in the northwestern WNP, and changes in the mean 25

480 position of TC genesis in the WNP and ENP). We demonstrated that the interdecadal 481 change in TC activity in the Pacific during the late 1990s was forced by the La 482 Niña-like SSTA pattern caused by the CRS in the Pacific during the middle to late 483 1990s. The numerical experiments further indicated that the declines in TC activity in 484 the eastern WNP and northern WSP were attributable to the negative SSTA in the CP. 485 Conversely, the increases in TC activity in the western WNP and southern WSP were 486 caused by the K-shaped warm SST pattern in the western Pacific. Although the 487 number of TCs in the ENP does not show significant interdecadal change, the mean 488 TC genesis mosition shifted westward significantly from 95 W to 102 W during IP2. 489 The numerical experiments also indicated that the westward shift of the mean TC 490 genesis position in the ENP and WNP was caused primarily by the large-scale 491 low-level cyclonic and anticyclonic circulation anomalies in the ENP and WNP, 492 respectively. 493 Figure 2 shows that the mean positions of TC genesis in WNP and ENP 494 experienced a significant westward shift from IP1 to IP2. The major reason the 495 change in TC tracks in the ENP did not exhibit a westward shift as detected in the 496 WNP (Fig. 3a, 3b) is that the anomalous low-level cyclone resulting from the La 497 Niña-associated negative SSTAs (Fig. 9c) weakened the climatologically westward 498 steering flow (Fig. 11b), substantially preventing the westward movement of TC 26

499 tracks in the ENP. 500 Notably, the westward shift of the mean TC genesis position in the WNP was 501 accompanied by an increase in TC activity in the western WNP and southeastern 502 South China Sea (e.g., Yang et al. 2012). Previous studies have shown that the 503 increase in TC activity in the western WNP could lead to a significant increase in TC 504 landfalls in Taiwan (Tu et al. 2011; He et al. 2015). Irrespective of the increase in TC 505 genesis frequency, our study showed that the modulation effect of the CRS on the 506 steering flow (i.e., westward extension and northward curving near the Philippine Sea) 507 also contributed substantially to the increase in TC landfalls in Taiwan after the late 508 1990s. The numerical experiments showed that the increase in TC genesis frequency 509 in the western WNP resulted from the K-shaped warm SST induced thermodynamic 510 changes. In other words, the K-shaped warm SST created wet conditions in the 511 mid-troposphere via the anomalous vertical vapor transport (Figs. 6 and 10), which 512 favored TC genesis. Furthermore, the K-shaped warm SST modified the steering flow 513 (westerly), causing it to extend substantially westward (Fig. 11b). The numerical 514 experiments showed consistency with the observational analysis, indicating that the 515 modulation effect of the K-shaped warm SST on the steering flows and TC genesis 516 frequency might have led to the significant increase in TC landfalls in Taiwan and 517 southeastern China after the late 1990s (Figs. 3 and 11). 27

518 Another notable finding concerns the increase in TC landfalls in northeastern and 519 northwestern Australia after the late 1990s. Although the SVD analysis (Fig. 7) 520 showed that the interdecadal change of SVD2 was independent of global warming, the 521 warming trend enhanced the magnitude of PC2 by approximately 20%. The increase 522 in TC genesis in the WSP occurred off the coast of northeastern and northwestern 523 Australia, where rapid warming after the 1980s was identified (not shown). Because 524 the increase in TC genesis frequency in these regions was caused primarily by 525 changes in the thermodynamics (i.e., warm SSTs), the question of whether TC 526 landfalls in northeastern and northwestern Australia will become more frequent in 527 response to global warming is under investigation. 528 In this study, we focused on the influence of the CRS on the interdecadal change in 529 TC activity in the Pacific Basin. The impact of the CRS during 1976 1977 (Fig. 7, 530 left panels) was not considered. Figure 1c illustrates that the number of TCs in the 531 ENP also increased significantly during 1976 1977. However, this interdecadal 532 change was not observed in either the WNP or WSP (Fig. 1b, 1c). Regarding the lack 533 of impact of the CRS on TC activity in the WNP for 1976 1977, Fig. 7 shows that 534 PC1 and PC2 were out of phase during the period of inactive TC genesis during 535 1976 1984, whereas they were in phase during 1998 2011. This observation suggests 536 that the large-scale environmental change associated with SVD1 in 1976 1977 was 28

537 partially offset by the effect of SVD2. In addition to the large-scale factors, the 538 interdecadal variation in TC activity in the WNP was also affected by other processes, 539 such as intraseasonal oscillations (Hsu et al. 2008; Kim et al. 2008; Ha et al. 2014), 540 which cannot be measured using the interdecadal time scale. The interdecadal change 541 in TC activity in other periods, such as the increase in TC activity in the ENP during 542 1976 1977 and the increase in TC activity in the WNP during the late 1980s, might be 543 attributable to other processes not currently under investigation. 544 545 546 Acknowledgements 547 Valuable suggestions of two reviewers are highly appreciated. This study was 548 supported by NSC 102-2111-M845-002 and NSC 103-2111-M845-002. TL 549 acknowledges support by ONR grant N000141210450 and by the International Pacific 550 Research Center that is sponsored by the Japan Agency for Marine-Earth Science and 551 Technology (JAMSTEC). This is SOEST contribution number xxxx and IPRC 552 contribution number xxx. The authors thank the ICTP group for providing the 553 SPEDDY model. 554 555 29

556 References 557 558 559 Bracco A., F. Kucharski, F. Molteni, W. Hazeleger and C. Severijns, 2005: Internal and forced modes of variability in the Indian Ocean. Geophys. Res. Lett., 32, L12707. 560 Callaghan, J., and S. B. Power., 2011: Variability and decline in the number of severe 561 tropical cyclones making land-fall over eastern Australia since the late nineteenth 562 century. Clim Dyn 37: 647-662. 563 Camargo, S. J., and A. H. Sobel, 2005: Western North Pacific tropical cyclone intensity 564 and ENSO, J. Climate, 18, 2996-3006. 565 Camargo, S.J., K.A. Emanuel and A.H. Sobel, 2007. Use of a genesis potential index 566 to diagnose ENSO effects on tropical cyclone genesis. J. Climate, 20, 567 4819-4834. 568 Chan, J. C. L., 2000: Tropical cyclone activity over the western North Pacific 569 associated with El Niño and La Niña events, J. Climate, 13, 2960-2972. 570 Chan, J. C. L. 2005: Interannual and interdecadal variations of tropical cyclone 571 activity over the western North Pacific. Meteorol. Atmos. Phys., 89, 143 152, doi 572 10.1007/s00703-005-0126-y. 573 Chia, H. H. and C. F. Ropelewski, 2002: The interannual variability in the genesis 574 location of tropical cyclones in the northwest Pacific, J. Climate, 15, 2934-2944. 30

575 Choi, Y., K. J. Ha, C. H. Ho, C.E. Chung, 2015,:Interdecadal change in typhoon 576 genesis condition over the western North Pacific, Clim. Dyn., 45, 3243-3255. 577 Emanuel, K. A., and D. S. Nolan, 2004: Tropical cyclone activity and global climate. 578 Preprints, 26 th Conf. on Hurricanes and Tropical Meteorology, Miami, FL, Amer. 579 Meteor. Soc., 240 241. 580 Gill, A. E., 1980: Some simple solutions for heat-induced tropical circulation. Quart. 581 J. Roy. Meteor. Soc., 106, 447-462. 582 Ha, Y., Z. Zhong, Y. Sun, and W. Lu, 2014: Decadal change of South China Sea 583 tropical cyclone activity in mid-1990s and its possible linkage with intraseasonal 584 variability, J. Geophys. Res., 119, 5331 5344, doi:10.1002/2013jd021286. 585 He, H., J Yang, D. Gong, R. Mao, Y. Wang, M. Gao, Decadal changes in tropical 586 cyclone activity over the western North Pacific in the late 1990s, Clim. Dyn Online 587 publication date: 11-Mar-2015. 588 Hong, C-C., Y.-K. Wu, T. Li, and C.-C. Chang, 2013: The climate regime shift in 589 Pacific during 1996/1997., doi: 10.1007/s00382-013-1867-9 590 Hsu, H.-H., Y.-L. Chen, A.-K. Lo, C.-H. Hung, W.-S. Kau, and C.-C. Wu, 2008: 591 Intraseasonal Oscillation-Tropical Cyclone Coupling in the Western North Pacific 592 during the 2004 Typhoon Season. In Recent Progress in Atmospheric Sciences: 593 Applications to the Asia-Pacific Region, Ed. K. N. Liou and M. D. Chou, World 31

594 Scientific, Singapore, 49-65. 595 Hsu, P.-C., P.-S. Chu, H. Murakami, and X. Zhao, 2014: An abrupt decrease in the 596 late-season typhoon activity over the Western North Pacific. J. Clim. 27, 597 4296 4312. doi: 10.1175/JCLI-D-13-00417.1 598 Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis 599 Project, Bull. Amer. Meteor. Soc., 77, 437 471. 600 Knapp, K. R., M. C. Kruk, D. H. Levinson, H. J. Diamond, and C. J. Neumann, 601 2010: The International Best Track Archive for Climate Stewardship (IBTrACS): 602 Unifying tropical cyclone best track data. Bulletin of the American Meteor. 603 Society, 91, 363-376., doi:10.1175/2009bams2755.1 604 Kim, J. H., C. H. Ho, H. S. Kim, C. H. Sui, and S. K. Park, 2008: Systematic 605 Variation of Summertime Tropical Cyclone Activity in the Western North Pacific 606 in Relation to the Madden Julian Oscillation. J. Climate, 21, 1171 1191. 607 doi: http://dx.doi.org/10.1175/2007jcli1493.1 608 Kucharski F., Molteni F, and Bracco A, 2006: Decadal interactions between the 609 western tropical Pacific and the North Atlantic Oscillation. Clim. Dyn. 26, 79-91 610 Li, T. 2012: Synoptic and climatic aspects of tropical cyclogenesis in Western North 611 Pacific. Nova Science Publishers, Inc., Eds. K. Oouchi and H. Fudeyasu, 612 Chap.3, pp.61-94. 32

613 Li, Z., W. Yu, T. Li, V. S. N. Murty, and F. Tangang, 2013: Bimodal character of 614 cyclone climatology in Bay of Bengal modulated by monsoon seasonal cycle, J. 615 Climate, 26 (3), 1033-1046. 616 Lin, I. I. and J. C. L. Chan, 2015: Recent decrease in typhoon destructive potential 617 and global warming implications. Nature Communications 6, 7182 618 Liu, K. S. and J. C. L, Chan, 2013: Inactive period of western North Pacific tropical 619 cyclone activity in 1998-2011., J. Climate, 26, 2614-2630. 620 Lupo, A. R., 2011: The Interannual and Interdecadal Variability in Hurricane Activity, 621 Recent Hurricane Research - Climate, Dynamics, and Societal Impacts, Prof. 622 Anthony Lupo (Ed.), ISBN: 978-953-307-238-8, InTech, Available from: 623 http://www.intechopen.com/books/recent-hurricane-research-climate-dynamics-a 624 nd-societal-impacts/the-interannual-and-interdecadal-variability-in-hurricane-acti 625 vity. 626 Matsuura, T., M. Yumoto, and S. Iizuka, 2003: A mechanism of interdecadal 627 variability of tropical cyclone activity over the western North Pacific. Climate 628 Dynamics, 21, 105 117, DOI 10.1007/s00382-003-0327-3. 629 Matthew, H., E, S. McGregor, P. Spence, G.A. Meehl, A. Timmermann, W. Cai, A. S. 630 Gupta, M. J. McPhaden, A. Purich, A. Santoso., 2014: Recent intensification of 631 wind-driven circulation in the Pacific and the ongoing warming hiatus. Nat. Clim. 33

632 Change 4. 222-227, DOI: 10.1038/nclimate2106 633 Meehl, G. A., Hu, A., Arblaster, J. M., Fasullo, J. Y. & Trenberth, K. E., 634 2013: Externally forced and internally generated decadal climate variability 635 associated with the Interdecadal Pacific Oscillation. J. Clim. 26, 7298 7310. 636 Molteni, F., 2003: Atmospheric simulations using a GCM with simplified 637 physical parametrizations. I. Model climatology and variability in multi-decadal 638 experiments. Clim. Dyn. 20, 175-191, DOI: 10.1007/s00382-002-0268-2. 639 Murakami, H., B.Wang, T. Li, and A. Kitoh, 2013: Projected increase in tropical 640 cyclones near Hawaii. Nature Climate Change 3, 749 754, 641 doi:10.1038/nclimate1890. 642 Ramsay, H. A., Camargo, S. J. & Kim, D., 2012 :Cluster analysis of tropical cyclone 643 tracks in the Southern Hemisphere. Clim. Dyn. 39, 897 917 644 Rayner, N. A., D. E. Parker, E. B. Horton, C. K. Folland, L. V. Alexander, D. P. 645 Rowell, E. C. Kent, and A. Kaplan, 2003: Global analyses of sea surface 646 temperature, sea ice, and night marine air temperature since the late nineteenth 647 century, J. Geophys. Res., 108, 4407, doi:10.1029/2002jd002670. 648 Rodionov, S. N., 2004: A sequential algorithm for testing climate regime shifts, 649 Geophys. Res. Lett., 31, L09204, doi:10.1029/2004gl019448 650 Salinger, M. J., S. McGree, F. Beucher, S. B. Power, F. Delage., 2014 :A new index 34

651 for variations in the position of the South Pacific convergence zone 652 1910/11 2011/2012. Clim. Dyn. 43, 881-892. 653 Tu, J.-Y., C. Chou, P. Huang and R.-H. Huang, 2011: An abrupt increase of intense 654 typhoons over the western North Pacific in early summer. Environmental 655 Research Letters. 6, 034013, doi: 10.1088 /1748-9326/6/3/034013 656 Wang, B., and J. C. L. Chan, 2002: How strong ENSO events affect tropical storm 657 activity over the western North Pacific, J. Climate, 15, 16431658. 658 Wu L., B. Wang, and S. Geng, 2005: Growing typhoon influence on East Asia. 659 Geophys. Res. Lett., 32, L18703, doi:10.1029/2005gl022937. 660 Wu, L., C. Wang, and B. Wang, 2015: Westward shift of western North Pacific 661 tropical cyclogenesis. Geophys. Res. Lett., 42, 1537-1542. 662 Yang, L., Y. Du, S. Xie, and D. Wang. 2012: An interdecadal change of tropical 663 cyclone activity in the South China Sea in the early 1990s. Chinese Journal of 664 Oceanology and Limnology, 30, 953-959. 665 35